Introduction and Corpus

Introduction

Jazz and Classical music are different. The main difference is probably because when you playing jazz, you have to improvise and compose on the spot and range things, however, with classical music, you are playing all the notes composed. That is, Jazz performers play freestyle music, while classical music performers play pre-structured music.

A study even proved that there is a different process occurred in the brains of the jazz and classical pianists. In particular, the jazz pianists’ brains began re-planning sooner than the classical pianists’ brains. Classical pianists tend to focus on ‘how’ to play the music. This means their focus is on technique and the personal expression they add to the piece. Jazz pianists, on the other hand, focus on ‘what’ to play, meaning they are always prepared to improvise and adapt the notes they’re playing.

2 music genre under 2 contrary mindsets are different. Thus,this storyboard is going to analyzing the difference between jazz and classical music trying to illustrate what is the major difference between such freestyle and disciplined music by graphs and to capiture how the mind set per genre influences the sound.

The analysis will first show the differences comparing 2 typical and popular songs from 2 music genre and then take a general look at the 2 genre samples.

Corpus

To completed the research focusing on the piano songs of both music genres ( acousticness>0.9 for all the tracks), 2 Spotify playlists will be taken. The most popular song from each album as representatives to do song specific analysis. 2 songs are: “Peace piece” by Bill Evans from the Jazz album, and “First love” by Yiruma from the Classical album.

Playlist Name Discription Number of tracks Popular song
Jazz Jazz Piano seleted by Jazz Pianoists Jazz piano pieces selected by pianists. Solo acoustic piano only; no rhythm or vocals. 65 Peace piece _ Bill Evans
Classical Classical Piano Essentials The world’s greatest piano songs, featuring Fur Elise, Clair De Lune, River flows in you etc.) 135 First love _ Yiruma

Spotify API have 12 features for a track, namely, key, mode, time_signature, acousticness, danceability, energy, instrumentalness, liveness, loudness, speechiness, valence, tempo. To see explanation for each variable, please click “Here

Chromagrams features of 2 songs


Chromagrams at the left show the intensity of the twelve pitch classes for each song per small fragment while the music is playing. One interesting fact is that “Peace piece” from the Jazz list prefers playing G a lot in the key of C major with relatively gentle harmony showed by the lighter blue in other pitch classes, while Yiruma seems to stick to the three major keys when playing a song in A major.

In addition, in “Peace piece”, the song is with a strong preference on G with some mention of B, A, E and C. However, the melody of “river flows in you” seems more “colorful” with B, A, F#, E, D and C#. No basis for any keys.

Chordgram - Peace piece by Bill Evans


The chordgram at the left shows the chords in “Peace piece”. The piece is segmented according to Spotify’s estimates, and the distances represented are cosine distances from Spotify’s chroma vectors of the keys in the y axis. The song is in C chords with A7 chords.

Chordgram - River flows in you by Yiruma


The chordgram at the left shows the chords dist in “River flows in you”. The piece is segmented according to Spotify’s estimates, and the distances represented are cosine distances from Spotify’s chroma vectors of the keys in the y axis. The whole song is on A chords.

Self-Similarity Matrices - Peace piece by Bill Evans


Self-similarity matrics graphically depict the similarity of song itself. One audio file is represented as a square. Each side of the square is proportional to the length of the piece, and time runs from left to right as well as from bottom to top. Similar regions are dark while dissimilar regions are the novelty which is illustrated by the yellow line. Thus there is always a dark diagonal line running from bottom left to top right because the audio is always the most similar to itself at any particular time. The Chroma and timbre grams do not show a huge difference is because both tracks are played only by piano. However, the music structure between 2 songs differs.

For the track “peace by piece”, the song suddenly changes after 2 mins and have another new section after 5 mins.It does not have a obvious repeated part after the intro.

Self-Similarity Matrices - River flows in you by Yiruma


By contrast, “River flows in you” seems to be more structured, reflecting by the regularsequential squares and yellow lines.

Classification Algorithms

           Truth
Prediction  Classical Jazz
  Classical       116   19
  Jazz             24   48


After 2 songs anaylsis, let’s, then, take a look from a general view.

The heatmap at the left shows the performance of a classifier trying to distinguish the jazz playlist from the classical playlist. It works with the whole sample set with 140 Classical songs and 67 Jazz songs.

It displays the results, after 10-fold cross-validation, of a nearest-neighbour classifier with 10 Spotify features (danceability, energy, loudness, speechiness, acousticness , instrumentalness, liveness, valence, tempo, durations) and 12 pitch classes of the song.

The performance of the classifier is above average: overall classification accuracy is 78%, with a J-index of 0.53. The precision of the classifier’s predictions of Jazz music is lower than the one predicts Classical.

4 major track differences between 2 list


Using a random-forest classifier, the most important features for classifying tracks among these playlists are:loudness, energy, valance.

This graph at the left, therefore, is made to shows the above 3 dimensions plus modes of Jazz and Classical music. The x axis shows ‘loudness’ and the y axis shows Spotify’s ‘valence’ feature. The major mode is in orange whereas the minor mode is in blue. The size of each dot represents the energy of each track.

In Jazz tracks, the proportion of majors ( 88% ) is relatively larger than the proportion of major in Classical music list (63%). Although Jazz modes are more or less different from the tradition scale, it seems like Spotify prefers to classifying them into major.

The loudness is more varied for Classical music. It is mainly distributed from -26 to -13 in Jazz list, but this range is larger in Classical music list where it is distributed from -36 to -15 with a higher standard deviation judging by the more scattered dots. Comparing with the relatively similar valence feature in the Classical list (except the highest and lowest valence tracks), Classical music is peaceful on average whereas Jazz music is evenly distributed across different energy. According to the dots size and the figures in Figure 2, Classical music is peaceful on average whereas Jazz music is evenly distributed across different energy.

All in all, the visualisation shows that Jazz music is (according to Spotify) rather louder and rather energetic than the Classical music. This may tell that as a freesytle music genre, Jazz players put more emphasizing on the role of melogy, rhythm and emotion.

Figure 1. A summary of the mean of each variable.

Genre M_danancibility M_enerergy M_loudness M_speechiness M_liveness M_valence M_tempo M_instrumentalness M_acousticness
Jazz 0.423 0.127 -20.6 0.0544 0.162 0.179 96.9 0.887 0.981
Classialc 0.376 0.0609 -27.0 0.0562 0.103 0.226 93.0 0.895 0.991

Figure 2. Energy distrybution (Vertical dotted line represents the average value of each genre.)

energy

energy

Hierarchical Clustering


The hierarchical clustering diagram on the left shows the similarities of the songs in each playlist. For convenient reading and comparing, I take the top 50 songs from each playlist to make the comparison, where the first graph shows the Jazz playlist and the second graph shows the Classical playlist.

Overall, there are more subsets in classical music and the clustering trees are higher. This explains that Jazz music is relatively more similar to each other. This probably implies that although Jazz is freestyle music, Jazz musicians are not completely ”free“. There might be some structure in their mind to playing the melody to make sure the next pop up key will not sound”wrong". The structure may be highly correlated to tempo or mode. However, Classical music, as pre-written music, can be revised over and over again before it officially released. It has the opportunity to try and correct until the composer is satisfied with his master piece.

Conclusion